CN104382582B - A kind of device that dynamic electrocardiogram (ECG) data is classified - Google Patents

Abstract

The invention discloses a kind of device that dynamic electrocardiogram (ECG) data is classified, including: electrocardio-data collection device；Ecg information data base；Electrocardiographicdata data acquisition device；Electrocardiographicdata data screening plant, for electrocardiographicdata data is carried out difference analysis, filters out the electrocardiographicdata data of significant difference；Feature combination acquisition device, at least two in the electrocardiographicdata data having significant difference filtering out electrocardiographicdata data screening plant carry out feature combination, obtain the combination of multiple feature；Grader screening plant, its multiple features combination using multiple grader to obtain feature combination acquisition device is tested, to filter out optimum classifier and optimal characteristics combination；Classification results output device, for receiving the personal information of patient and the pathological data relevant with cardiomotility state, and the optimum classifier filtered out according to grader screening plant carries out classification output category result with optimal characteristics combination to the pathological data relevant with cardiomotility state of patient.

Description

A kind of device that dynamic electrocardiogram (ECG) data is classified

Technical field

The present invention relates to heart disease medical field, in particular to one, dynamic electrocardiogram (ECG) data is carried out The device of classification.

Background technology

At present, the index patient having sudden death risk carrying out risk stratification clinically has left ventricular ejection to divide Number, heart rate variability, heart rate turbulence, heart rate decelerations power, T ripple electrical alternations etc., above majority refers to Mark can calculate from ambulatory electrocardiogram.But, these Risk Stratification Methods are difficult to exactly at present Risk of being died suddenly by height patient makes a distinction from low sudden death risk population, it is therefore desirable to integrate various technical side Sudden death risk patient is layered by method, grader be then to the sudden death key that is layered of risk patient because of Element, uses grader can classify the dynamic electrocardiogram (ECG) data of the patient having sudden death risk, then uses one Fixed method carries out layering to sorted dynamic electrocardiogram (ECG) data can die suddenly risk patient from low sudden death by height Risk population makes a distinction.

MUSTT tests (A randomized study of the prevention of sudden death in patients with coronary artery disease.N Engl J Med 1999；341:1882-1890) combine low Ejection fraction and intracardiac electrophysiology inspection confirm examination to go out high sudden death risk patient, and SCD-HeFT Test (Amiodarone or an implantable cardioverter-defibrillator for congestive heart failure.N Engl J Med 2005；352:225-237) confirm low ejection fraction can examination to go out applicable ICD pre- Anti-patient, ABCD tests (The ABCD (Alternans Before Cardioverter Defibrillator) Trial:strategies using T-wave alternans to improve efficiency of sudden cardiac death prevention.J Am Coll Cardiol.2009Feb 10；53 (6): 471-9) low ejection fraction is confirmed The accuracy of sudden death prediction can be significantly improved in conjunction with T ripple electrical alternations and intracardiac electrophysiology inspection.These are large-scale Test all confirms that the effectiveness of some risk stratification instruments and two or more instrument combine and can substantially carry The effectiveness of high stratification.But it has the drawback that some check is invasive inspection, and some are non-invasive Check that its accuracy is the highest, in conjunction with two or three inspections test only by simple " with or without " judge, Although improve risk stratification ability, but its sensitivity being the most not fully up to expectations.

Summary of the invention

The invention provides a kind of device that dynamic electrocardiogram (ECG) data is classified, dynamic in order to patient Electrocardiogram (ECG) data is classified.

For reaching above-mentioned purpose, the invention provides a kind of device that dynamic electrocardiogram (ECG) data is classified, Including:

Electrocardio-data collection device, for gathering the personal information of patient and relevant with cardiomotility state Pathological data；

Ecg information data base, for storing personal information and and the heart that electrocardio-data collection device gathers The pathological data that dirty active state is relevant；

Electrocardiographicdata data acquisition device, relevant with cardiomotility state in ecg information data base Pathological data processes, and obtains the electrocardiographicdata data that the pathological data relevant with cardiomotility state is corresponding；

Electrocardiographicdata data screening plant, carries out difference for the electrocardiographicdata data obtaining electrocardiographicdata data acquisition device Property analyze, filter out the electrocardiographicdata data of significant difference；

Feature combination acquisition device, for the significant difference that has filtering out electrocardiographicdata data screening plant At least two in electrocardiographicdata data carry out feature combination, obtain the combination of multiple feature；

Grader screening plant, it uses multiple spies that feature combination acquisition device is obtained by multiple grader Levy combination to test, to filter out optimum classifier and optimal characteristics combination；

Classification results output device, for receiving the personal information of patient and relevant with cardiomotility state Pathological data, and the optimum classifier that filters out according to grader screening plant and optimal characteristics combination The pathological data relevant with cardiomotility state of patient is carried out classification output category result.

Preferably, personal information includes sex, age and blood group, the pathology relevant with cardiomotility state Data include New York Heart Association, History of Coronary Heart Disease, myocardial infarction medical history, hypertension history, valve Sick medical history, congenital heart disease medical history, cardiomyopathy medical history, diabetes medical history, cerebrovascular medical history, smoking history, History of drinking history, cardiovascular diseases's family history, sudden death family history, motion exercise situation, heredopathia medical history, contraction Pressure, diastolic pressure, Body Mass Index, whether pacemaker, ICD treatment situation, catheter ablation situation, Coronary artery bypass grafting situation, coronary artery bracket situation, B receptor blocking agent behaviour in service, calcium-channel antagonists use Situation, ACEI/ARB behaviour in service, diuretic behaviour in service, antiarrhythmic drug use history, ocean When Radix Rehmanniae class drug use history, lipid lowerers use history, WENXIN KELI to use the detection of history, dynamic electrocardiogram to continue Between, total heart rate, the fastest heart rate, the slowest heart rate, average heart rate, room early sum, non-standing room speed, Lowns classification, average NN, SDNN, SDANN, ASDNN, rMSSD, pNN50, pNN50a, PNN50b, BB50, BB50a, BB50b, very low frequency, low frequency, high frequency, wideband frequency, Yong Hupin Rate, low high ratio, TO%, TSmm/RR, TD, CCTS, TFD, heart rate decelerations power, FQRS, RMS40, LAS, phase, QT/RR between microvolt T ripple electrical alternations MTWA, QT ms, QTds, QTc Slope, QTd/RRd, QT variability, QTVI, T crest end time, the P ripple time limit, P ripple dispersion, When J wave height, J ripple become alternately, J ripple dispersion, QRS time limit, chamber pass to retardance situation, Q ripple Limit, Q depth of convolution degree, C reflection albumen, super quick C reflection albumen, homocysteine, CHLO, TG, LDL, HDL, BNP, NT-proBNP, creatinine, left atrial diameter, LVED, Left ventricular posterior wall thickness, left interventricular septal thickness, Left Ventricular Ejection Fraction, FS, mitral incompetence, Tricuspid valve are anti- Stream, aortic regurgitation, interim Abnormal Wall Motion, LM, LAD, LCX, RCA, LM, LAD, LCX, TIMI classification, Internal-media thickness and Plaques score.

Preferably, electrocardiographicdata data acquisition device obtain electrocardiographicdata data at least include heart rate variability metrics, Heart rate decelerations power, heart rate acceleration and heart rate turbulence index, wherein,

Heart rate variability metrics includes heart rate variability time domain index and heart rate variability frequency-domain index, heart rate Variability time domain index include between RR between phase average, RR phase average stdev between phase standard deviation, RR, Phase standard deviation average and triangle index between RR, heart rate variability frequency-domain index includes general power, extremely low frequency merit Rate, low frequency power, extremely high frequency power and high frequency power,

Heart rate turbulence index includes turbulence onset and turbulence slope.

Preferably, appointing during grader is naive Bayesian, support vector machine and artificial neural network algorithm Meaning one.

Preferably, device is classified according to the probability of patient's heart sudden death risk, accordingly, and classification Device screening plant filters out optimum classifier in the following manner:

S1: according to the personal information and relevant with cardiomotility state stored in ecg information data base The risk of sudden cardiac death probability of the patient that pathological data is corresponding, is divided into high-risk sudden death risk patient by patient With low danger sudden death risk patient；

S2: combine the electrocardiographicdata data included according to each feature, be respectively adopted multiple grader pair and the heart The relevant pathological data of dirty active state is classified, by the grader the most close with the classification results in S1 As optimum classifier.

The device classifying dynamic electrocardiogram (ECG) data that the present invention provides combines multiple electrocardiographicdata data to dynamically Electrocardiogram (ECG) data is classified, and can be used for cardiac risk layering and sudden death probability layering, it is possible to pre-for sudden death Anti-and clinical decision provides guidance instruction, substantially increases the accuracy of electrocardiogram (ECG) data classification.

Accompanying drawing explanation

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to reality Execute the required accompanying drawing used in example or description of the prior art to be briefly described, it should be apparent that below, Accompanying drawing in description is only some embodiments of the present invention, for those of ordinary skill in the art, On the premise of not paying creative work, it is also possible to obtain other accompanying drawing according to these accompanying drawings.

Fig. 1 is the structural representation of the device classifying dynamic electrocardiogram (ECG) data of one embodiment of the invention Figure；

Fig. 2 is a RR interval series figure；

Fig. 3 is the spectrogram that the RR interval series figure shown in Fig. 2 is corresponding；

Fig. 4 is the ambulatory electrocardiogram of a patient；

Fig. 5 is the RR interval series figure that the ambulatory electrocardiogram shown in Fig. 4 is corresponding；

Fig. 6 is to be labelled with heart rate decelerations cardiac cycle and the RR interval series figure of heart rate acceleration cardiac cycle；

Fig. 7 is the RR interval series figure having divided heart rate section；

Fig. 8 is the RR interval series figure after the mutually whole sequence in position；

The meansigma methods schematic diagram in the corresponding cycle that Fig. 9 calculates；

Figure 10 is phase sequence number schematic diagram between sinus rate and RR after ventricular premature contraction；

Figure 11 is the ROC curve of algorithm of support vector machine.

Description of reference numerals: 1-electrocardio-data collection device；2-ecg information data base；3-electrocardiographicdata data obtains Fetching is put；4-electrocardiographicdata data screening plant；5-feature combination acquisition device；6-grader screening plant；7- Classification results output device.

Detailed description of the invention

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out Clearly and completely describe, it is clear that described embodiment is only a part of embodiment of the present invention, and It is not all, of embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art are not paying Go out the every other embodiment obtained under creative work premise, broadly fall into the scope of protection of the invention.

Fig. 1 is the structural representation of the device classifying dynamic electrocardiogram (ECG) data of one embodiment of the invention Figure, as it can be seen, the device classifying dynamic electrocardiogram (ECG) data that the present invention provides includes electrocardiogram (ECG) data Harvester 1, ecg information data base 2, electrocardiographicdata data acquisition device 3, electrocardiographicdata data screening plant 4, Feature combination acquisition device 5, grader screening plant 6 and classification results output device 7, wherein,

Electrocardio-data collection device 1 is for gathering the personal information of patient and relevant with cardiomotility state Pathological data, utilize electrocardio-data collection device 1 by the gatherer process standardization of dynamic electrocardiogram (ECG) data, Standardization, improves the collecting flowchart of dynamic electrocardiogram (ECG) data.In the present embodiment, electrocardio-data collection device 1 personal information gathered includes sex, age and blood group, electrocardio-data collection device 1 gathers and the heart The relevant pathological data of dirty active state includes that New York Heart Association, History of Coronary Heart Disease, myocardial infarction are sick History, hypertension history, valvular heart disease medical history, congenital heart disease medical history, cardiomyopathy medical history, diabetes medical history, brain Angiopathy medical history, smoking history, history of drinking history, cardiovascular diseases's family history, sudden death family history, motion exercise shape Condition, heredopathia medical history, shrink pressure, whether diastolic pressure, Body Mass Index, pacemaker, ICD treat Situation, catheter ablation situation, coronary artery bypass grafting situation, coronary artery bracket situation, B receptor blocking agent use shape Condition, calcium-channel antagonists behaviour in service, ACEI/ARB behaviour in service, diuretic behaviour in service, the anti-heart Rule arrhythmic agents use history, digitaloid drugs use history, lipid lowerers use history, WENXIN KELI use history, Dynamic electrocardiogram detection persistent period, total heart rate, the fastest heart rate, the slowest heart rate, average heart rate, room are early Sum, non-standing room speed, Lowns classification, average NN, SDNN, SDANN, ASDNN, RMSSD, pNN50, pNN50a, pNN50b, BB50, BB50a, BB50b, very low frequency, low Frequently, high frequency, wideband frequency, user's frequency, low high ratio, TO%, TSmm/RR, TD, CCTS, TFD, heart rate decelerations power, FQRS, RMS40, LAS, microvolt T ripple electrical alternations MTWA, QT ms, Between QTds, QTc when phase, QT/RR slope, QTd/RRd, QT variability, QTVI, T crest end Between, P ripple time limit, P ripple dispersion, J wave height, J ripple become alternately, J ripple dispersion, the QRS time limit, Chamber pass to retardance situation, the Q ripple time limit, Q depth of convolution degree, C reflection albumen, super quick C reflection albumen, Homocysteine, CHLO, TG, LDL, HDL, BNP, NT-proBNP, creatinine, Left atrial diameter, LVED, Left ventricular posterior wall thickness, left interventricular septal thickness, Left Ventricular Ejection Fraction, FS, mitral incompetence, tricuspid regurgitation, aortic regurgitation, interim Abnormal Wall Motion, LM, LAD, LCX, RCA, LM, LAD, LCX, TIMI classification, Internal-media thickness and Plaques score.

Personal information that ecg information data base 2 gathers for storing electrocardio-data collection device 1 and with The pathological data that cardiomotility state is relevant, wherein, ecg information data base 2 may utilize internet site or Other share means alternately, form the ecg information data base that the world is shared.

Relevant with cardiomotility state in ecg information data base 2 of electrocardiographicdata data acquisition device 3 Pathological data process, refer to the electrocardio obtaining the pathological data relevant with cardiomotility state corresponding Mark.In the present embodiment, the electrocardiographicdata data that electrocardiographicdata data acquisition device 3 obtains at least includes heart rate variability Property index, heart rate decelerations power, heart rate acceleration and heart rate turbulence index, wherein,

Heart rate variability (HRV) refers to the Micro-fluctuations of R--R interval between successive heartbeat, that reflects the heart Dirty sympathetic nerve and the tonicity of vagal activity and harmony, be a kind of to detect Autonomic nerve block Noninvasive index.Heart rate variability metrics includes heart rate variability time domain index and heart rate variability frequency domain Index, wherein, heart rate variability time domain index includes between RR phase mark between phase average (MEAN), RR Accurate phase standard deviation average (ASDNN) between phase average stdev (SDANN), RR between poor (SDNN), RR With triangle index (Ti), it is illustrated in figure 2 a RR interval series figure, each heart rate variability time domain index root According to RR interval series figure, calculated by below equation (1)～(4) respectively:

$\mathrm{MEAN}=\stackrel{\‾}{\mathrm{RR}}=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{RR}}_i/N---\left(1\right)$

$\mathrm{SDNN}=\sqrt{\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{({\mathrm{RR}}_i-\stackrel{\‾}{\mathrm{RR}})}^2}---\left(2\right)$

$\mathrm{SDANN}=\sqrt{\underset{i=1}{\overset{288}{\mathrm{\Σ}}}{(\stackrel{\‾}{R_i}-\stackrel{\‾}{{\mathrm{RR}}_{5\min }})}^2/288}---\left(3\right)$

$\mathrm{ASDNN}=\underset{i=1}{\overset{288}{\mathrm{\Σ}}}{\mathrm{SDNNindex}}_i/288---\left(4\right)$

Heart rate variability frequency-domain index by the spectrogram that RR interval series figure is corresponding is analyzed obtain, Fig. 3 is the spectrogram that the RR interval series figure shown in Fig. 2 is corresponding, as it can be seen, heart rate variability frequency Territory index includes general power (0.0033～0.4Hz), extremely low frequency power (0.0033～0.04Hz), low Frequently power (0.04～0.15Hz), extremely high frequency power (> 0.4Hz) and high frequency power (0.15～0.4Hz).

Heart rate decelerations power and heart rate acceleration are extracted by following steps:

Step 1: extract RR interval series figure, be illustrated in figure 4 the ambulatory electrocardiogram of a patient, Fig. 5 It show the RR interval series figure that the ambulatory electrocardiogram shown in Fig. 4 is corresponding；By the dynamic electrocardiogram of 24 hours Figure (Fig. 4) is converted into heartbeat serial number abscissa, with the value of phase between cardiac cycle i.e. RR as vertical coordinate Sequence chart (Fig. 5).

Step 2: detection labelling add (subtracting) speed cycle: by each cardiac cycle value (RR_i) with this week Previous cardiac cycle (the RR of phase_i-1) compare, determine that this cycle belongs to heart rate decelerations cardiac cycle Or heart rate accelerates cardiac cycle, then with different symbols in addition labelling.It is illustrated in figure 6 and is labelled with the heart Rate deceleration cardiac cycle and heart rate accelerate the RR interval series figure of cardiac cycle, as it can be seen, ratio is previous The cardiac cycle that individual cardiac cycle is big is defined as deceleration periods, labels it as " * " in Fig. 6；Ratio is previous The cardiac cycle that individual cardiac cycle is little is defined as the acceleration cycle, labels it as " o " in Fig. 6.

Step 3: determine heart rate section, heart rate section used when carrying out " the mutually whole sequence in position " refers to slow down with each Point or acceleration point be when being heart rate section center, and fetch bit is on the left of deceleration point or acceleration point and right side aroused in interest respectively The number in cycle, and the concrete numerical value that left and right respectively takes how many cardiac cycles needs to refer to minimal heart rate.Such as figure 7 show the RR interval series figure having divided heart rate section, wherein, V₁、V₂、V₃、V₄It is four hearts Rate section, it can be seen that the present embodiment is when carrying out heart rate decelerations power and analyzing, and heart rate hop count value is set to 30 Between the phase, then when this means that centered by selected deceleration point, respectively take 15 aroused in interest Cycle one heart rate section of composition.

Step 4: the mutually whole sequence in position, centered by selected deceleration point (" * " point), carries out not concentric rate section Superposition, be illustrated in figure 8 the RR interval series figure after the mutually whole sequence in position.

Step 5: calculate the average period of corresponding sequence number, after " the mutually whole sequence in position ", calculates the corresponding cycle respectively Meansigma methods:

(1) X (0): the meansigma methods of phase between the RR of all center position；

(2) X (1): the meansigma methods of first cardiac cycle on the right side of central point；

(3) X (-1): the meansigma methods of first cardiac cycle on the left of central point；

(4) X (-2): the meansigma methods of second all cardiac cycle on the left of central point.

Be illustrated in figure 9 the meansigma methods schematic diagram in the corresponding cycle calculated, as seen from the figure X (0), X (1), X (-1), the size of X (-2).

Step 6: calculate heart rate decelerations power, after calculating the average of X (0), X (1), X (-1), X (-2) respectively, Result is substituted into following formula (5) again and carries out calculating heart rate decelerations power DC that i.e. can get:

$\mathrm{DC}=\frac{X\left(0\right)+X\left(1\right)-X(-1)-X(-2)}{4}---\left(5\right)$

It addition, in step 4 and step 5 to above-mentioned signal processing time, marked the heart the most simultaneously Rate acceleration point (" o " point), and use same flow process, substitute into acceleration point and the acceleration cycle just can calculate the heart 4 averages (Z (0), Z (1), Z (-1), Z (-2)) that rate acceleration AC is relevant, substitute into following formula (6) i.e. Can calculate the value of heart rate acceleration AC of person under inspection:

$\mathrm{AC}=\frac{Z\left(0\right)+Z\left(1\right)-Z(-1)-Z(-2)}{4}---\left(6\right)$

Heart rate turbulence index includes turbulence onset and turbulence slope.

After what turbulence onset (TO) described is ventricular premature contraction, whether sinus rhythm exists the phenomenon of acceleration.As Figure 10 show after ventricular premature contraction phase sequence number schematic diagram between sinus rate and RR, and the calculating of turbulence onset is public Formula is phase (phase between after also referred to as) between the RR with front 2 sinus rhythms of compensatory after date of ventricular premature contraction With, deduct between the RR of 2 sinus rhythms before ventricular premature contraction coupling interval the phase (phase between before also referred to as) With, both differences are again divided by the latter, and the result of gained is TO.In Fig. 10, ventricular premature contraction is compensatory Between after date front 2 sinus rhythms RR between the phase be respectively R1 and R2, before ventricular premature contraction coupling interval 2 sinus rhythms RR between the phase be respectively R_-1And R_-2, it is calculated as follows out turbulence onset TO:

$\mathrm{TO}=\frac{(R_1+R_2)-(R_{-1}+R_{-2})}{R_{-1}+R_{-2}}\×100\%$

Turbulence slope (TS) is whether to there is sinus rhythm deceleration phenomenon after quantitative analysis ventricular premature contraction.

The calculation procedure of turbulence slope TS is as follows:

Step 1: time value between the RR of front 20 sinus rhythms after sensing chamber's early appearance, and these RR Between the value of phase as vertical coordinate, using the heartbeat sequence number of phase between RR as abscissa, draw time value between RR Sequence chart；

Step 2: between RR in the sequence chart of time value, the sinus rhythm to every 5 continuous print heartbeat sequence numbers Point makes the regression line, and wherein forward greatest gradient is TS.

Electrocardiographicdata data screening plant 4 carries out difference for the electrocardiographicdata data obtaining electrocardiographicdata data acquisition device Property analyze, filter out the electrocardiographicdata data of significant difference；

Feature combination acquisition device 5, for having significant difference to what electrocardiographicdata data screening plant filtered out Electrocardiographicdata data at least two carry out feature combination, obtain multiple feature combination；

Grader screening plant 6 uses multiple features that feature combination acquisition device is obtained by multiple grader Combination is tested, and to filter out optimum classifier and optimal characteristics combination, wherein, grader can be Any one in naive Bayesian, support vector machine and artificial neural network algorithm.

In the present embodiment, the device classifying dynamic electrocardiogram (ECG) data that the present invention provides is for according to trouble The probability of person's risk of sudden cardiac death is classified, and accordingly, grader screening plant is in the following manner Filter out optimum classifier:

S1: according to the personal information and relevant with cardiomotility state stored in ecg information data base The risk of sudden cardiac death probability of the patient that pathological data is corresponding, is divided into high-risk sudden death risk patient by patient With low danger sudden death risk patient；

S2: combine the electrocardiographicdata data included according to each feature, be respectively adopted multiple grader pair and the heart The relevant pathological data of dirty active state is classified, by the grader the most close with the classification results in S1 As optimum classifier.In this course, carry out Feature Selection by automatization and manual method, will The electrocardiographicdata data not having obvious characteristic of division is rejected, and remainder has the electrocardiographicdata data of obvious characteristic of division retain And carry out feature combined test.It follows that different feature combinations is entered with different classifier algorithms respectively Row machine learning, after using leaving-one method and cross-validation method to process, obtains different classifier algorithm not Classification efficiency value with feature combination.Then, the optimal classification of each grader is combined and extracts, And carry out the ratio of area (AUC) value under sensitivity, specificity, accuracy rate and experimenter's operating curve Relatively, and then filter out there is the optimal synthesis classification grader of usefulness and characteristic of division combination.

Classification results output device 7 is for receiving the personal information of patient and relevant with cardiomotility state Pathological data, and the optimum classifier that filters out according to grader screening plant and optimal characteristics combination The pathological data relevant with cardiomotility state of patient is carried out classification output category result.

The accuracy classified dynamic electrocardiogram (ECG) data for the checking present invention, inventor acquires 208 Position patient dynamic electrocardiogram (ECG) data, average 28 months follow up a case by regular visits to work after, it is determined that dynamic electrocardiogram Classification results.By the analysis to ecg information, extract 10 electrocardios having significant difference altogether and refer to Mark, carries out feature combination for feature combination acquisition device, tests through feature selection, leaving-one method and intersection After demonstration processes, the positive findings of 90% can be identified, and LVEF can only identify about 20%, The AUC of support vector machine reaches as high as 0.8902, and SDNN and AUC that single index is the highest It is only 0.78.More than experiment proves that the present invention has bigger in heart disease dynamic electrocardiogram (ECG) data classification field Advantage, classification accuracy is higher.

It is the ROC curve (experimenter's operating function curve) of algorithm of support vector machine as shown in figure 11, Algorithm of support vector machine is used to carry out dynamic electrocardiogram (ECG) data is classified in the present invention, its result such as figure Shown in 11.This area under curve can reach 0.89, and between left ventricular ejection fraction (LVEF), RR Phase standard deviation (SDNN) and heart rate decelerations power (DC) are the most significant lower.

The device classifying dynamic electrocardiogram (ECG) data that the present invention provides combines multiple electrocardiographicdata data to dynamically Electrocardiogram (ECG) data is classified, and can be used for cardiac risk layering and sudden death probability layering, it is possible to pre-for sudden death Anti-and clinical decision provides guidance instruction, substantially increases the accuracy of electrocardiogram (ECG) data classification.

One of ordinary skill in the art will appreciate that: accompanying drawing is the schematic diagram of an embodiment, in accompanying drawing Module or flow process not necessarily implement necessary to the present invention.

One of ordinary skill in the art will appreciate that: the module in device in embodiment can be according to enforcement Example describes in the device being distributed in embodiment, it is also possible to carries out respective change and is disposed other than the present embodiment In one or more devices.The module of above-described embodiment can merge into a module, it is also possible to further Split into multiple submodule.

Last it is noted that above example is only in order to illustrate technical scheme, rather than to it Limit；Although the present invention being described in detail with reference to previous embodiment, the ordinary skill of this area Personnel it is understood that the technical scheme described in previous embodiment still can be modified by it, or Wherein portion of techniques feature is carried out equivalent；And these amendments or replacement, do not make relevant art The essence of scheme departs from the spirit and scope of embodiment of the present invention technical scheme.

Claims (4)

1. the device that dynamic electrocardiogram (ECG) data is classified, it is characterised in that including:

Electrocardio-data collection device, for gathering the personal information of patient and relevant with cardiomotility state Pathological data；

Ecg information data base, for store described electrocardio-data collection device gather personal information and The pathological data relevant with cardiomotility state；

Electrocardiographicdata data acquisition device, for having with cardiomotility state in described ecg information data base The pathological data closed processes, and the electrocardio obtaining the pathological data relevant with cardiomotility state corresponding refers to Mark；

Electrocardiographicdata data screening plant, is carried out for the electrocardiographicdata data obtaining described electrocardiographicdata data acquisition device Difference analysis, filters out the electrocardiographicdata data of significant difference；

Feature combination acquisition device, for having significance poor to what described electrocardiographicdata data screening plant filtered out Different at least two in electrocardiographicdata data carry out feature combination, obtain the combination of multiple feature；

Grader screening plant, it uses multiple grader to obtain many to described feature combination acquisition device The combination of individual described feature is tested, to filter out optimum classifier and optimal characteristics combination；

Classification results output device, for receiving the personal information of patient and relevant with cardiomotility state Pathological data, and the optimum classifier that filters out according to described grader screening plant and optimal characteristics Combine the pathological data relevant with cardiomotility state to patient and carry out classification output category result；

Wherein, described personal information includes sex, age and blood group, described relevant with cardiomotility state Pathological data include New York Heart Association, History of Coronary Heart Disease, myocardial infarction medical history, hypertension history, Valvular heart disease medical history, congenital heart disease medical history, cardiomyopathy medical history, diabetes medical history, cerebrovascular medical history, smoking History, history of drinking history, cardiovascular diseases's family history, sudden death family history, motion exercise situation, heredopathia medical history, Shrink pressure, whether diastolic pressure, Body Mass Index, pacemaker, ICD treat situation, catheter ablation shape Condition, coronary artery bypass grafting situation, coronary artery bracket situation, B receptor blocking agent behaviour in service, calcium-channel antagonists Behaviour in service, ACEI/ARB behaviour in service, diuretic behaviour in service, antiarrhythmic drug use history, Digitaloid drugs uses history, lipid lowerers to use history, WENXIN KELI to use the detection of history, dynamic electrocardiogram to continue Time, total heart rate, the fastest heart rate, the slowest heart rate, average heart rate, room early sum, non-standing room Speed, Lowns classification, average NN, SDNN, SDANN, ASDNN, rMSSD, pNN50, PNN50a, pNN50b, BB50, BB50a, BB50b, very low frequency, low frequency, high frequency, broadband frequency Rate, user's frequency, low high ratio, TO%, TSmm/RR, TD, CCTS, TFD, heart rate decelerations Power, FQRS, RMS40, LAS, microvolt T ripple electrical alternations MTWA, QT ms, QTds, QTc Between phase, QT/RR slope, QTd/RRd, QT variability, QTVI, T crest end time, P ripple time Limit, P ripple dispersion, J wave height, J ripple become alternately, J ripple dispersion, QRS time limit, chamber pass to Retardance situation, Q ripple time limit, Q depth of convolution degree, C reflection albumen, super quick C reflect albumen, homotype half Guang In propylhomoserin, CHLO, TG, LDL, HDL, BNP, NT-proBNP, creatinine, left room Footpath, LVED, Left ventricular posterior wall thickness, left interventricular septal thickness, Left Ventricular Ejection Fraction, FS, Mitral incompetence, tricuspid regurgitation, aortic regurgitation, interim Abnormal Wall Motion, LM, LAD, LCX, RCA, LM, LAD, LCX, TIMI classification, Internal-media thickness and Plaques score.

The device that dynamic electrocardiogram (ECG) data is classified the most according to claim 1, it is characterised in that Wherein, the electrocardiographicdata data that described electrocardiographicdata data acquisition device obtains at least includes heart rate variability metrics, the heart Rate decelerative force, heart rate acceleration and heart rate turbulence index, wherein,

Described heart rate variability metrics includes heart rate variability time domain index and heart rate variability frequency-domain index, Described heart rate variability time domain index includes between RR between phase average, RR between phase standard deviation, RR that the phase is average Value is phase standard deviation average and triangle index between standard deviation, RR, and described heart rate variability frequency-domain index includes always Power, extremely low frequency power, low frequency power, extremely high frequency power and high frequency power,

Described heart rate turbulence index includes turbulence onset and turbulence slope.

The device that dynamic electrocardiogram (ECG) data is classified the most according to claim 1, it is characterised in that Described grader is any one in naive Bayesian, support vector machine and artificial neural network algorithm.

The device that dynamic electrocardiogram (ECG) data is classified the most according to claim 1, it is characterised in that Described device is classified according to the probability of patient's heart sudden death risk, and accordingly, described grader sieves Screening device filters out optimum classifier in the following manner:

S1: according to the personal information stored in described ecg information data base and have with cardiomotility state The risk of sudden cardiac death probability of patient corresponding to pathological data closed, is divided into high-risk sudden death risk by patient Patient and low danger sudden death risk patient；

S2: combine the electrocardiographicdata data included according to each feature, be respectively adopted multiple grader to described The pathological data relevant with cardiomotility state is classified, and the most close with the classification results in S1 is divided Class device is as optimum classifier.